Claude Sonnet 5 Matches Opus Performance at Half the Price Through August

The traditional gap between Anthropic’s mid-tier and premium model lines is narrowing, and the pricing structure around that convergence tells the more interesting story.

Claude Sonnet 5 launches today with performance Anthropic describes as similar to Opus 4.8, its previous premium tier. That comparison matters because Opus models have historically carried the better agentic capabilities and the higher price tags to match. Sonnet 5 is priced at $2 per million input tokens and $10 per million output tokens, at least through August 31, when those rates rise to $3 and $15 respectively.

The agentic improvements are where Anthropic is placing its emphasis. Sonnet 5 can complete complex tasks that its predecessor, Sonnet 4.6, could not finish at all, and it checks its own output without being prompted to do so. Anthropic also says it shows lower rates of hallucination and sycophancy, and is better at refusing malicious requests.

A Promotional Window Built Into the Launch

The introductory pricing through August 31 is a notable structural choice. Anthropic is effectively subsidizing adoption during the initial rollout period before reverting to rates that still undercut Opus but carry a 50 percent markup on inputs. The model is available across all plans and is now the default for Free and Pro users, which means Anthropic is also using Sonnet 5 to raise its baseline product without charging more upfront.

Sonnet 5 improves over Sonnet 4.6 across reasoning, tool use, coding, and knowledge work. Opus models retain the edge in agentic capabilities overall, though the gap has visibly compressed. Whether that remaining gap justifies the Opus price premium is now a harder case for Anthropic to make to developers evaluating both tiers.